COURSE NOTES: HOMOLOGICAL ALGEBRA

COURSE NOTES: HOMOLOGICAL ALGEBRA AMNON YEKUTIELI Contents 1. Introduction 1 2. Categories 2 3. Free Modules 13 4. Functors 18 5. Natural Transformations 23 6. Equivalence of Categories 29 7. Opposite Rings and Tensor Products 32 References 37 1. Introduction Lecture 1, 7 March 2018 Motivating discussion and history: proving the Brouwer Fixed Point Theorem, using the homology functors H i : Top Ab, and the structure of the abelian groups H i (B n ) and H i (S n 1 ). This is in the handwritten pages. What do we keep from this topological story? Functors. Complexes and their (co)homology. But the categories that will interest us will be the category of left A-modules Mod A over a ring A, and its relatives. The functors we will work with will be additive functors F : Mod A Mod B. For such a functor F we will study its left derived functors L i F and its right derived functors R i F. Here is a sample result on the structure of modules over a commutative ring A that we will prove in the course. This is a theorem that the methods of the course Commutative Algebra could not produce for us. Theorem 1.1. Let A be a noetherian commutative ring, and let M be a finitely generated A-module. The two conditions below are equivalent. (i) M is flat. 1

(ii) M is projective. All the concepts above, with the exception of projective module, were studied in the previous course. These concepts shall be explained in our course when the time comes. The proof of the theorem relies on the derived functors and Tor A i (, ) = L i( A ) Ext i A (, ) = Ri Hom A (, ). Their role is analogous to the role that the homology groups played in the proof of the Brouwer Theorem. 2. Categories Definition 2.1. A category C is a mathematical structure consisting of these ingredients: A set Ob(C), whose elements are called the objects of C. For every pair C, D Ob(C) there is a set Hom C (C, D), whose elements are called the morphisms from C to D, and are denoted by f : C D. For every triple C, D, E Ob(C) there is a function Hom C (D, E) Hom C (C, D) Hom C (C, E), (д, f ) д f called composition. For every C Ob(C) there is a morphism id C Hom C (C,C) called the identity morphism. There are two axioms: (Associativity) For composable morphisms f,д,h there is equality h (д f ) = (h д) f. (Identity) For a morphism f : C D there is equality f = f id C = id D f. Remark 2.2. In order to avoid set-theoretic difficulties, we assume that there is a given set U called a universe. The universe U is large enough so as to contain as elements all the mathematical structures that will concern us (rings, groups, topological spaces) and their cartesian products, power sets, and so on. A set S will be called a small set if S U. Our standing assumption is that every category C satisfies Ob(C) U, and that Hom C (C, D) U for every C, D Ob(C). We will not discuss set theoretical issues beyond the above. More details on this foundational aspect can be found in [Mac2, Section I.6]. Often we write C C as a shortcut for C Ob(C). We now list a few important examples of categories. Example 2.3. The category Set has as objects all the small sets. Thus Ob(Set) = U. The morphisms f : S T in Set are the functions; composition of morphisms is the usual composition of functions; and id S (s) = s for every s S. Example 2.4. The category Grp has as objects all groups. (Only the small groups, i.e. their underlying sets must be small.) The morphisms ϕ : G H in Grp are the group homomorphisms. Compositions and identity morphisms are the usual ones. 2

Example 2.5. The category Ab has as objects all abelian groups. The morphisms ϕ : M N in Ab are the group homomorphisms. Compositions and identity morphisms are the usual ones. Example 2.6. The category Ring has as objects all rings. The morphisms f : A B in Ring are the ring homomorphisms. (A ring homomorphism f is required to respect unit elements: f (1 A ) = 1 B.) Compositions and identity morphisms are the usual ones. Example 2.7. The category Ring c has as objects all commutative rings. The morphisms are the ring homomorphisms. Compositions and identity morphisms are the usual ones. Let A be a ring. Recall that an element a A is called central if a b = b a for all b A. The center of A is the set of all central elements in it, and we denote it by Cent(A). It is easy to see that Cent(A) is a subring of A, and moreover Cent(A) is a commutative ring. There is a similar story for groups. The next exercise will help us understand the center better. Exercise 2.8. Find a ring homomorphism f : A B such that f (Cent(A)) is not contained in Cent(B). Do this also for groups. Definition 2.9. Let K be a commutative ring. (1) A K-ring is a pair (A, f A ) consisting of a ring A and a ring homomorphism f A : K A. The ring homomorphism f A is called the structural homomorphism of A, and usually it will be kept implicit, and we shall just say that A is a K-ring. (2) Suppose (A, f A ) and (B, f B ) are K-rings. A K-ring homomorphism д : (A, f A ) (B, f B ) is a ring homomorphism д : A B such that д f A = f B. I.e. the diagram K f A A д f B in Ring is commutative. (3) A central K-ring is a K-ring (A, f A ) such that f A (K) Cent(A). (4) The category of central K-rings is the category Ring/ c K, whose objects are the central K-rings (item (3)), and whose morphism are the K-ring homomorphisms (item (2)). Exercise 2.10. Show that every ring A admits a unique ring homomorphism Z A, and this makes A into a central Z-ring. Conclude that B Ring/ c Z = Ring. So far our examples involved large categories (i.e. they had many objects). This is not always the case: Example 2.11. Let A be a ring. We define a category A as follows. There is a single object x, so Ob(A) = {x}. The set of morphisms is Hom A (x, x) := A. Composition is multiplication in A, and id x := 1 A. 3

Definition 2.12. Let A be a ring. A left A-module is an abelian group M, together with a function A M M, (a,m) a m called multiplication, satisfying these conditions for all a, b A and m, n M : Associativity: (a b) m = a (b m). Distributivity: (a + b) m = (a m) + (b m) and a (m + n) = (a m) + (a n). Unit: 1 A m = m. Exercise 2.13. Let A be a ring. (1) Define what is a right A-module (relying on Definition 2.12). (2) Assume A is a commutative ring. Show that there is no difference between left and right A-modules. Definition 2.14. Let A be a ring. Given left A-modules M and N, an A-linear homomorphism ϕ : M N is a homomorphism of abelian groups such that ϕ(a m) = a ϕ(m) for all a A and m M. The set of A-linear homomorphism ϕ : M N is denoted by Hom A (M, N ). Convention 2.15. From here on we fix a nonzero commutative base ring K. We assume by default that all rings are central K-rings, and all ring homomorphisms are over K. Given a ring A, by default A-modules are left A-modules. Definition 2.16. Let A be a ring. The category Mod A has the (left) A-modules as its objects, and the morphisms are the A-linear homomorphisms. Thus Hom Mod A (M, N ) = Hom A (M, N ). Exercise 2.17. Let A be a central K-ring and let M, N Mod A. Show that Hom A (M, N ) has a canonical K-module structure. Exercise 2.18. Let K be a commutative ring, A a central K-ring and M a left A-module. Define End A (M) := Hom A (M, M). (1) Show that End A (M) is a ring, in which multiplication is composition. (2) Show that the function K End A (M), λ λ id M, is a ring homomorphism, and moreover it makes End A (M) into a central K-ring. The next exercise says that the converse is also true: Exercise 2.19. Let K be a commutative ring, A a central K-ring and M a K-module. Suppose д : A End K (M) is a K-ring homomorphism. For a A and m M define a m := д(a)(m). Prove that this makes M into a left A-module. The discussion of K-linear categories, that was started yesterday, will resume next week. 4

Lecture 2, 14 March 2018 Before continuing with the material, I want to say a few words on the direction the course will take. The some of the students attended the course Commutative Algebra last semester, and there they learned a lot about categories and functors. Other students may not have seen this material before, and thus they need to be introduced to it gradually and effectively. In order not to bore the first group of students, and still be accessible to the second group, I have decided to concentrate on the noncommutative aspects of categories and functors. In particular, we will learn: Right B-modules and the ring B op ; A-B-bimodules and the ring A K B op. The tensor product M A N for a right A-module M and a left A-module N. A baby case of Morita equivalence, between Mod A and Mod B, where A is a ring and B := Mat r (A), the ring of r r matrices for r 1. This will be done with concrete formulas. (We may do the general Morita equivalence later.) Now to the material. Let M, N, P Mod K. Recall that a function β : M N P is called K-bilinear if it s additive in the two arguments: β(m 1 + m 2, n) = β(m 1, n) + β(m 2, n), β(m, n 1 + n 2 ) = β(m, n 1 ) + β(m, n 2 ) and respects multiplication by elements of K : β(λ m, n) = β(m, λ n) = λ β(m, n) for all m,m 1,m 2 M, all n, n 1, n 2 N and all λ K. Definition 2.20. Let K be a nonzero commutative ring. A K-linear category is a category M, together with a K-module structure on each of the morphism sets Hom M (M, N ). The condition is that for every triple of objects L, M, N M the composition function is K-bilinear. Hom M (M, N ) Hom M (L, M) Hom M (L, N ) Proposition 2.21. If A is a central K-ring, then Mod A is a K-linear category. Proof. Let s write M := Mod A. We start by specifying the K-module structure on each of the morphism sets Hom M (M, N ). (This was Exercise 2.17.) The zero element of Hom M (M, N ) is the zero homomorphism 0 : M N, i.e. the constant function m 0. Given A-linear homomorphisms ϕ,ψ : M N, their sum ϕ + ψ is the A-linear homomorphism (ϕ + ψ )(m) := ϕ(m) + ψ (m). Given an A-linear homomorphism ϕ : M N and an element λ K, let λ ϕ : M N be the homomorphism (λ ϕ)(m) := λ ϕ(m). 5

We must check that λ ϕ is A-linear, and this is where we use the fact that A is a central K-ring. Take an element a A. Then (λ ϕ)(a m) = λ ϕ(a m) = λ a ϕ(m) = a λ ϕ(m) = a (λ ϕ)(m) for every m M. We now have to prove that composition in M is K-bilinear. Let ϕ : L M and ψ : M N be morphisms in M. For every λ K and l L we have Therefore ((λ ϕ) ψ )(l) = (λ ϕ)(ψ (l)) = λ ϕ(ψ (l)) = λ (ϕ ψ )(l) = (λ (ϕ ψ ))(l). (λ ϕ) ψ = λ (ϕ ψ ). The remainder of the proof (checking the other three equations) is left as an exercise. Exercise 2.22. Finish the proof of the last proposition, by verifying that and ϕ (λ ψ ) = λ (ϕ ψ ) (ϕ 1 + ϕ 2 ) ψ = (ϕ 1 ψ ) + (ϕ 2 ψ ) ϕ (ψ 1 + ψ 2 ) = (ϕ ψ 1 ) + (ϕ ψ 2 ). Most categories that we talked about do not admit a group structure on their morphisms sets; at least not in any natural way. The next two exercises explore this point. Exercise 2.23. Consider the category of groups Grp. For a pair of groups G, H the set Hom Grp (G, H) has a special element, namely the constant homomorphism ϕ 1 : G H, ϕ 1 (д) := 1 H. You might think that there s a group structure on Hom Grp (G, H), in which ϕ 1 is the unit element. Show that this is false, by calculating ϕ 1 ϕ 1 for nonabelian G and H (e.g. take both to be S 3 ). Exercise 2.24. Consider the category of rings Ring. Find a pair of rings A, B such that the set Hom Ring (A, B) is empty. So it can t be a group. Exercise 2.25. This exercise reverses Proposition 2.21. Let M be a K-linear category, and let M M. Show that End M (M) := Hom M (M, M) is a central K-ring. We now leave linear categories for a while. Definition 2.26. Let C be a category. A subcategory C of C consists of a subset Ob(C ) Ob(C), and for each pair of objects C, D Ob(C ) a subset (2.27) Hom C (C, D) Hom C (C, D). There is no condition on the subset Ob(C ). There are two conditions on the morphism subsets Hom C (C, D): (Identities) id C Hom C (C,C) for every C Ob(C ). (Closure under composition) If C, D, E Ob(C ), f Hom C (D, E), then д f Hom C (C, E). We write C C. Hom C (C, D) and д 6

Of course C on its own is a category, with the operations inherited from C. Definition 2.28. A subcategory C C is called a full subcategory if for every pair of objects C, D Ob(C ) there is equality Hom C (C, D) = Hom C (C, D). In other words, a full subcategory C C is determined by selecting a subset Ob(C ) of Ob(C). This subset could be empty (not interesting), finite, etc. Example 2.29. The category Ab fin of finite abelian groups is a full subcategory of Ab. Example 2.30. Let us choose one group from each isomorphisms class of finite abelian groups. This gives us a countable subset S Ob(Ab fin ). Let C Ab fin be the full subcategory such that Ob(C) = S. In some sense everything happens already inside C. This observation will be made precise later. Example 2.31. Consider the category of rings Ring. Let C be the following subcategory of Ring : it has all the objects, but the the morphisms f : A B in C are the those ring homomorphisms that satisfy f (Cent(A)) Cent(B). It is easy to see that C is indeed a subcategory. Exercise 2.8 shows that there are less morphisms in C. So this is not a full subcategory. Example 2.32. Here s a variation on Example 2.11. Let A A be a subring. Define the subcategory A A with Ob(A ) = {x} and This is not a full subcategory. Hom A (x, x) := A. Definition 2.33. Let C be a category. A morphism f : C D in C is called an isomorphism if there is a morphism д : D C such that f д = id C and д f = id D. If an exercise is labeled optional (like the next one), then you should solve it only if you think it is not trivial. In any case, do not submit it in writing. Exercise 2.34. (Optional) If f : C D is an isomorphism, then the morphism д in the definition above is unique. The morphism д in Definition 2.33 is called the inverse of f, and is denoted by f 1. Definition 2.35. A category G is called a groupoid if all the morphisms in it are isomorphisms. Exercise 2.36. Let C be a category. Show that there is a subcategory C of C, that has all the objects, and whose morphisms are the isomorphisms in C. Show that C is a groupoid. Here is an important special case. Exercise 2.37. Let M be a K-linear category, and consider the subcategory M M as in the previous exercise. Take an object M M. We know from Exercise 2.25 that A := End M (M) is a central K-ring. Show that End M (M) is the group A of invertible elements of the ring A. 7

Course Notes Amnon Yekutieli 25 April 2018 Definition 2.38. Let C be a category. (1) An object I C is called an initial object if for every object C C there is a unique morphism I C. (2) An object T C is called a terminal object if for every object C C there is a unique morphism C T. Exercise 2.39. Let C be a category. (1) If I and I are both initial objects of C, then there is a unique isomorphism I I. (2) If T and T are both terminal objects of C, then there is a unique isomorphism T T. Exercise 2.40. (1) Find initial and terminal objects in the category Set. (2) Find initial and terminal objects in the category Grp. (3) Find initial and terminal objects in the category Ring. (4) Find initial and terminal objects in the category Mod A for a ring A. We end this section with a discussion of products and coproducts. Let C be a category. By a collection of objects of C, indexed by a set I, we mean a function γ : I Ob(C). We usually denote this collection by {C i } i I, where C i := γ (i) Ob(C). Definition 2.41. Let C be a category, and let {C i } i I be a collection of objects of C. A product of this collection is a pair ( C, {pi } i I ) where C is an object of C, and {p i } i I is a collection of morphisms p i : C C i in C, called projections. The pair ( C, {p i } i I ) should have the following universal property: (P) Given an object D C, and a collection {f i } i I of morphisms f i : D C i, there exists a unique morphism f : D C such that f i = p i f. The notation for the object C is i I C i. And usually we leave the morphisms p i implicit. Here is a depiction of Definition 2.41 where I = {1, 2, 3}. C D f C p 1 p 2 p 3 C 1 C 2 C 3 f 1 f 2 f 3 C 1 C 2 C 3 A product is unique: Proposition 2.42. Let C be a category and let {C i } i I be a collection of objects of C. Suppose that ( ) ( C, {p i } i I and C, {p i } ) i I are both products of this collection of objects. Then there is a unique isomorphism д : C C such that p i д = p i for all i. 8

Exercise 2.43. Prove Proposition 2.42. From now on we say the product. Example 2.44. In the category Set all products exist; they are the usual cartesian products, with the usual projections on the coordinates. Exercise 2.45. Let A be a ring and let {M i } i I be a collection of objects of Mod A. Prove that the product ( M, {p i } i I ) in Mod A exists. (Hint: take the product M := i I M i in Set, and show that the set M has an A-module structure, such the projections p i are homomorphisms.) Exercise 2.46. Show that the category Ab fin of finite abelian groups has finite products (i.e. products indexed by finite sets I), but not infinite products. (Hints: for the first part use Exercise 2.45. For the second part find a concrete counterexample.) Solution 2.47 (of Exercise 2.46). Suppose {M i } i I is a collection of nontrivial finite groups, indexed by an infinite set I. We can t argue that the product in Ab is infinite, and thus it does not belong to Ab fin ; see Remark 2.58. Here is a correct proof. Let I be the set of positive integers, and for every i I choose a finite abelian group M i of size i (e.g. the cyclic group Z/(i)). Assume that the product ( ) M, {p i } i I in Abfin exists. For every i consider the collection of homomorphisms {ϕ i, j } j I, ϕ i, j : M j M i, defined as follows: { idmi if j = i ϕ i, j := 0 if j i. By the universal property we get a homomorphism ϕ : M M i, and p i ϕ = ϕ i,i = id Mi. Hence p i : M M i is surjective. This implies M M i = i. So M is infinite. Contradiction. 9

Course Notes Amnon Yekutieli 25 April 2018 Lecture 3, 21 March 2018 Here is the notion dual to product. Definition 2.48. Let C be a category, and let {C i } i I be a collection of objects of C. A coproduct of this collection is a pair ( C, {ei } i I ) where C is an object of C, and {e i } i I is a collection of morphisms e i : C i C. This pair ( C, {ei } i I ) should have the following universal property: (C) Given an object D C, and a collection {д i } i I of morphisms д i : C i D, there exists a unique morphism д : C D such that The notation for the object C is i I C i. д i = д e i. Here is a depiction of Definition 2.48 where I = {1, 2, 3}. C D д C C 1 e 1 e 2 C 2 e 3 C 3 д 1 C 1 д 2 C 2 д 3 C 3 Exercise 2.49. Let C be a category and let {C i } i I be a collection of objects of C. Suppose that ( ) ( C, {e i } i I and C, {e i } ) i I are both coproducts of the collection of objects {Ci } i I. Prove that there is a unique isomorphism h : C C such that e i = h e i. Example 2.50. In the category Set all coproducts exist; they are the disjoint union. Given a collection of sets {S i } i I, the functions e j : S j are the inclusions. Example 2.51. In the category Grp all coproducts exist, but they are very nasty. Given a finite collection of sets {G i } i I, indexed by I = {1,..., n}, the coproduct is the group i I G i := G 1 G n. Even if the G i are abelian, the coproduct isn t (unless all but one of the G i are trivial). For instance, if G i = Z, then G 1 G n is the free group on n generators. It very nonabelian (for n > 1): its center is trivial. For an infinite indexing set I, the coproduct is the union: G i = G j, i I J I j J where J runs over the finite subsets of I. i I S i 11

Example 2.52. In the category Ring c /K of commutative K-rings (see Examples 2.7 and 2.9) the coproduct is the tensor product. This was studied in the previous course (see [Ye2, Lecture 5, page 52]). Given a finite collection {A i } i I in Ring c /K, indexed by I = {1,..., n}, the coproduct is the ring A i := A 1 K... K A n. i I For instance, if A i = K[t i ], the polynomial ring in a variable t i, then A 1 K K A n = K[t 1,..., t n ], the commutative polynomial ring in n variables. For an infinite indexing set I, the coproduct is the union: A i = A j, i I J I j J where J runs over the finite subsets of I. Example 2.53. In the category Ring/ c K of central K-rings (see Example 2.9) the coproduct exists, and it is as nasty as in Example 2.51. For instance, if I = {1,..., n}, and A i = K[t i ], the commutative polynomial ring in a variable t i, then A i = K t 1,..., t n, i I the noncommutative polynomial ring in n variables. If n > 1 then the center is K. In a K-linear category M we usually say direct sum instead of coproduct, and write M i := M i. i I Exercise 2.54. Let A be a ring and let {M i } i I be a collection of objects of Mod A. Describe the coproduct, or direct sum, i I M i. (Hint: it is a submodule of the product.) Exercise 2.55. Prove that Ab fin does not have infinite direct sums (i.e. coproducts). (Warning: recall the comment on Exercise 2.46, and read Remark 2.58 below.) K-linear categories were introduced in Definition 2.20. Proposition 2.56. Let M be a K-linear category, and let {M i } i I be a collection of objects of M, indexed by a finite set I. Assume that the direct sum ( M, {e i } i I ) of the collection {M i } i I exists in M. Then: (1) The object M is also the product of the collection {M i } i I. Namely there are morphisms p i : M M i, such that the pair ( M, {p i } i I ) is the product of the collection {M i } i I. (2) The collections of morphisms {e i } i I and {p i } i I satisfy and i I p i e i = id Mi e i p i = id M. This proposition applies to M = Mod A for a central K-ring A. i I 12

Exercise 2.57. Prove Proposition 2.56. (Hint: use the universal property (C) of the coproduct to construct the morphisms p i.) If you find it easier, prove the proposition for M = Mod A, A Ring/ c K. Remark 2.58. Regarding Exercise 2.46, I said that it could happen that for a full subcategory C C, products and coproducts might be different in each category. This can be easily seen for the coproduct of abelian groups: given nontrivial abelian groups G 1 and G 2, their coproduct in Ab is not their coproduct in Grp, even though Ab Grp is a full subcategory. Indeed, the coproduct G 1 G 2 in Grp is nonabelian. 3. Free Modules We now leave the abstraction of categories for a while, to talk about an important concrete construction. In this section A is some nonzero ring (possibly noncommutative). Recall that modules are left modules, by default. Definition 3.1. Let M be an A-module and let X be a set. (1) The support of a function f : X M is the set Supp(f ) := {x X f (x) 0} X. (2) We denote by F fin (X, M) the set of finitely supported functions f : X M. (3) Given f F fin (X, M), its sum f (x) M is defined to be ( ) x X x X f (x) := x Supp(f ) f (x). Note that the second sum in ( ) is finite, so we are not doing anything illegal. Exercise 3.2. Let M be a module over the ring A. We know that the set F(X, M) of all functions f : X M is an A-module, by pointwise operations. Show that is an A-submodule. F fin (X, M) F(X, M) As before, a function f : X M can be considered as a collection {m x } x X of element of M, where m x := f (x). Under the guise of a collection, we sometimes use the notation m := {m x } x X. Thus m = f, two ways of referring to the same thing. 13

Lecture 4, 28 March 2018 Recall K is a nonzero commutative base ring, A is some nonzero central K-ring (possibly noncommutative), and all modules are by default left modules. Definition 3.3. Let M be an A-module. Given a collection m := {m x } x X of elements of M, and a finitely supported collection a := {a x } x X of elements of A, both indexed by the same indexing set X, the collection {a x m x } x X is a finitely supported collection of elements of M. Thus the sum a m := a x m x M exists. x X Definition 3.4. Let M be an A-module and let m := {m x } x X be a collection of elements of M; i.e. m F(X, M). (1) We say that the collection m generates the module M if for every element m M there exists some a F fin (X, A) such that m = a m. (2) We say that the collection m is linearly independent if the only element a F fin (X, A) such that a m = 0 is a = 0. (3) We say that the collection m is a basis of the module M if it generates M and it is linearly independent. Definition 3.5. An A-module M is called free if it has a basis. For a set X and an element x X we denote by δ x : X A the function { 1 if y = x (3.6) δ x (y) := 0 if y x. The support of δ x is {x}. Proposition 3.7. Let X be a set. Consider the collection δ := {δ x } x X of elements of F fin (X, A). (1) For every a F fin (X, A) there is equality a = a δ in F fin (X, A). (2) The A-module F fin (X, A) is free, with basis δ. Exercise 3.8. Prove this proposition. Example 3.9. Suppose the indexing set is X = {1,..., r } for some natural number r. Then The standard basis of A r is, in our notation, F fin (X, A) = F(X, A) = A r. δ = (δ 1,..., δ r ). 15

Proposition 3.10. Let M be an A-module, and letm := {m x } x X be a collection of elements of M. The function ϕ m : F fin (X, A) M, ϕ m (a) := a m is an A-module homomorphism. Moreover, it is the unique A-module homomorphism such that ϕ(δ x ) = m x for all x X. Exercise 3.11. Prove this proposition. ϕ : F fin (X, A) M Proposition 3.12. Let M be an A-module, and letm := {m x } x X be a collection of elements of M. Consider the homomorphism ϕ m : F fin (X, A) M. (1) m generates M iff ϕ m is surjective. (2) m is linearly independent iff ϕ m is injective. (3) m is a basis of M iff ϕ m is bijective. Exercise 3.13. Prove this proposition. We see that: Corollary 3.14. An A module M is free iff M F fin (X, A) for some set X. Proposition 3.15. Let f : X Y be a function between sets. There is a unique A-module homomorphism tr f : F fin (X, A) F fin (Y, A) such that for every x X. tr f (δ x ) = δ f (x) Proof. We give two proofs. First: use Proposition 3.10 with M := F fin (Y, A), and with the collection of elements m = {m x } x X defined by m x := δ f (x). Then tr f := ϕ m works. Second proof: for ψ F fin (X, A) define tr f (ψ )(y) := ψ (x) A. x f 1 (y) Remark 3.16. We can put the discrete topologies on the sets X and Y. The corresponding Lebesgue, or Borel, measures are the atomic measures that give each point the measure 1. The integrable functions X A and Y A are then the finitely supported functions. These are also the compactly supported functions. The homomorphism tr f : F fin (X, A) F fin (Y, A) is then the standard pushforward (or integration on fibers) tr f = of measures, or of compactly supported functions (or of distributions). See any (probably nonexistent) modern textbook on measure theory; or any modern textbook on analysis (e.g. [KaSc]). f 16

Here are two theorems about commutative rings, that were proved in the previous course. Theorem 3.17. Let K be a field. Every K-module M has a basis. See [Ye2, Lecture 3, page 28]. The proof relies om Zorn s Lemma (i.e. the Axiom of Choice). Theorem 3.18. Let A be a nonzero commutative ring, and let M be a free A-module. If m = {m x } x X and m = {m x } x X are two bases of M, then the sets X and X have the same cardinality. See [Ye2, Lecture 3, page 32]. The proof also relies on the Axiom of Choice (via the fact that a nonzero commutative ring A has a maximal ideal m). In the situation of Theorem 3.18, the cardinality of a basis of M is called the rank of M, and it is denoted by rank A (M). Example 3.19. Theorem 3.18 is false in general for noncommutative rings. Here is a counterexample. Let K be a field, and let V := F fin (N, K), which is a free K-module of countable rank. Define the central K-ring A := End K (V ). We can view an element v V as a column with finitely many nonzero entries. Then an element a A becomes an N N matrix with entries in K, that has finitely supported columns (i.e. the columns of a are elements of V ). The action of A on V is by matrix multiplication. λ 0,0 λ 0,1 µ 0 a v = λ 1,0 λ 1,1 µ 1...... Consider the free A-module M := A. Let M 0 M be the submodule consisting of the matrices a in which only the even numbered columns can be nonzero; and let M 1 M be the submodule consisting of the matrices a in which only the odd numbered columns can be nonzero. Then M = M 0 M 1 as left A-modules. On the other hand M 0 M 2 M as left A-modules. We see that A A 2 as left A-modules. This shows that free A-modules do not have well-defined ranks. The next few definitions and results (up to Exercise 3.27, about left noetherian rings) were not in the lecture. Definition 3.20. Let A be a ring. An A-module M is called a noetherian module if every submodule M M is finitely generated. Proposition 3.21. Let A be a ring and M an A-module. The following two conditions are equivalent: (i) M is a noetherian module. (ii) M satisfies the ascending chain condition: if M 0 M 1 M 1 is a chain of submodules of M, then M i = M i+1 for i 0. Exercise 3.22. Prove this proposition. 17

Definition 3.23. A ring A is called a left noetherian ring if A is a noetherian left module module over itself; i.e. if every left ideal a A is finitely generated as a left A-module. Theorem 3.24. Let A be a ring. The following two conditions are equivalent: (i) A is a left noetherian ring. (ii) Every finitely generated A-module M is a noetherian module. Exercise 3.25. Prove Theorem 3.24. (Hint: See the commutative case in [Ye2, Lecture 10, page 99].) Exercise 3.26. Let A be a nonzero left noetherian ring. Show that Theorem 3.18 holds for free A-modules. (Hint: study Example 3.19.) Exercise 3.27. Let A be a ring for which there exists a nonzero ring homomorphism f : A B to a commutative ring B. Then Theorem 3.18 holds for free A-modules. (Hint: modify the proof of Theorem 3.18, see [Ye2, Lecture 3, page 32].) Exercise 3.28. Let X be a set and M an A-module. Consider the collection of modules {M x } x X with M x := M. Show that F fin (X, M) = and x X X X M x F(X, M) = M x. 4. Functors Definition 4.1. Let C and D be categories. A functor consists of these ingredients: A function F : C D F Ob : Ob(C) Ob(D). For every pair C 1,C 2 Ob(C), a function There are two conditions: F C1,C 2 : Hom C (C 1,C 2 ) Hom D ( F(C1 ), F(C 2 ) ). f 1 f 2 (Composition) For all composable morphisms C 1 C 2 C 3 in C there is equality F C2,C 3 (f 2 ) F C1,C 2 (f 1 ) = F C1,C 3 (f 2 f 1 ) of morphisms F Ob (C 1 ) F Ob (C 3 ) in D. (Identity) For every object C of C there is equality F C,C (id C ) = id FOb (C) of morphisms from F Ob (C) F Ob (C) in D. Definition 4.2. Abbreviations: we will usually write F instead of F Ob and F C1,C 2, since this will be clear from the context. 18

Course Notes Amnon Yekutieli 25 April 2018 Example 4.3. Let A be a central K-ring. Given an A-module M, let F(M) be its underlying K-module. The resulting functor F : Mod A Mod K is called a forgetful functor, because it forgets some structure. We can also consider the underlying set G(M) of the module M. The resulting functor G : Mod A Set is also a forgetful functor. The functor G is a composition of functors (see definition below): where is this forgetful functor. G = H F, H : Mod K Set Definition 4.4. Let C, D and E be categories, and let F : C D and G : D E be functors. The composed functor G F : C E has ingredients (G F) Ob : Ob(C) Ob(E), and (G F) Ob := G Ob F Ob (G F) C1,C 2 : Hom C (C 1,C 2 ) Hom E ( (G F)Ob (C 1 ), (G F) Ob (C 2 ) ), (G F) C1,C 2 := G FOb (C 1 ),F Ob (C 2 ) F C1,C 2. The relevant diagram is (4.5) C F D G E G F The material from here until Exercise 4.8 (more example of functors) was not in the lecture. Exercise 4.6. Verify that G F is indeed a functor. Exercise 4.7. We know (by Exercise 2.8) that the center is not a functor from Grp to Ab. However, given a group G, let [G,G] be the subgroup of G generated by the commutators [д,h] := д h д 1 h 1. (1) Show that [G,G] is a normal subgroup. (2) Show that Ab(G) := G/[G,G] is an abelian group. (3) Show that Ab : Grp Ab is a functor. It is called the abelianization functor. 19

(4) Try to find the relation between the functor Ab and the embedding functor Emb : Ab Grp. (Recall that Ab is a full subcategory of Grp). Exercise 4.8. Let A be a nonzero ring. For a set X let Free(X) := F fin (X, A) Mod A. To a function f : X Y between sets we assign the A-module homomorphism from Proposition 3.15. (1) Prove that Free(f ) := tr f : F fin (X, A) F fin (Y, A) Free : Set Mod A is a functor. (2) Try to find the relation between the functor Free and the forgetful functor Forg : Mod A Set. The relation between the functors in Exercises 4.7 and 4.8 is called adjunction. We will discuss it later. 20

Lecture 5, 11 April 2018. Version 2 of notes. Definition 4.9. Let M and N be K-linear categories. A functor F : M N is called a K-linear functor if for every pair of objects M 1, M 2 M the function F : Hom M (M 1, M 2 ) Hom N ( F(M1 ), F(M 2 ) ) is K-linear. If K = Z then F is just called a linear functor, or an additive functor. Exercise 4.10. Let f : A B be a K-ring homomorphism. A B-module N can be made into an A-module by a n := f (a) n for a A and n N. Show that the formula above gives rise to a K-linear functor Rest f : Mod B Mod A. It is called the restriction functor corresponding to f. We sometimes write Rest B/A := Rest f. Note that the forgetful functor F from Example 4.3 is also a restriction functor. Exercise 4.11. Let A be a central K-ring. We know that Mod A is a K-linear category. Fix an A-module M. Thus for each N Mod A we have a K-module F M (N ) := Hom A (M, N ), and each homomorphism ϕ : N N in Mod A induces a homomorphism (1) Show that F M (ϕ) := Hom(id M, ϕ) : F(N ) F(N ). F M : Mod A Mod K is a K-linear functor. (2) Assume that A is a commutative ring. Shows that the functor F from item (1) can be upgraded to an A-linear functor such that F M : Mod A Mod A, F M = Rest A/K F M. Remark 4.12. Suppose that in the situation of Exercise 4.11 we were to define G M (N ) := Hom A (N, M) for N Mod A. Then G M (N ) belongs to Mod K. But for a homomorphism ϕ : N N, the homomorphism G M (ϕ) := Hom(ϕ, id M ) goes in the wrong direction. We will see soon that G M is a contravariant K-linear functor. This is a confusing, yet unavoidable, feature of category theory. We will study contravariant functors, as well as opposite categories, later just before talking about adjoint functors. Exercise 4.13. Let F : M N and G : N P be K-linear functors between K-linear categories. Show that G F is a K-linear functor. 21

Exercise 4.14. Let M and N be K-linear categories, and let F : M N be a K-linear functor. We know already that for every object M M the set of endomorphisms End M (M) is a central K-ring. Show that for every M M the function is a K-ring homomorphism. F : End M (M) End N (F(M)) Left A-modules were defined in Definition 2.12, and you were supposed to define right A-modules in Exercise 2.13. The next definition combines the previous two in a rather complicated way, that requires getting used to. Recall that rings are by default K-central. Definition 4.15. Let A and B be rings. A K-central A-B-bimodule is a K-module M, equipped with a left A-module structure and a right B-module structure, that commute with each other and respect the given K-module structure of M. Namely and for all m M, a A, b B and λ K. (a m) b = a (m b) (λ 1 A ) m = λ m = m (λ 1 B ) Example 4.16. Let A be a nonzero ring. Fix positive integers r and s. Define the rings A A B := Mat r (A) =....... A A and C := Mat s (A). Define the K-module M := Mat s r (A). Matrix multiplication gives a left action of C on M: and also a right action of B on M: (c,m) c m, (m,b) m b. The usual calculation in the linear algebra course shows that (c m) b = c (m b), regardless of the fact that the ring A is noncommutative. We see that M is a C-B-bimodule. another Exercise 4.17. Let M be an A-B-bimodule. (1) Show that for every N Mod A, the K-module Hom A (M, N ) has a B-module structure, with formula (b ϕ)(m) := ϕ(m b) for ϕ Hom A (M, N ), b B and m M. (2) Show that in this way is a K-linear functor. Hom A (M, ) : Mod A Mod B 22

Course Notes Amnon Yekutieli 25 April 2018 Remark 4.18. Suppose that s = 1 in the previous example, so that C = A and M is an A-B-bimodule. Later in the course (I hope) we will see that the functor Hom A (M, ) : Mod A Mod B is an equivalence of K-linear categories. This is what I call baby Morita Equivalence. Exercise 4.19. Let M and N be K-linear categories, and let F : M N be a K-linear functor. Pick a pair of objects M 1, M 2 M, and define A i := End M (M i ) and B i := End N (F(N i )). We know that F : A i B i are ring homomorphisms (by Exercise 4.14), and that there are restriction functors Rest Bi /A i : Mod B i Mod A i. (Exercise 4.10). Show that: (1) The K-module Hom M (M 1, M 2 ) is a K-central A 2 -A 1 -bimodule. (2) Show that F : Hom M (M 1, M 2 ) Hom N ( F(M1 ), F(M 2 ) ) is a homomorphism of A 2 -A 1 -bimodules. Convention 4.20. From here on we assume by default that all linear categories are K- linear, and all linear functors are K-linear. Also we assume that all bimodules are K- central. Of course if K = Z then the convention above is automatic. comment: To here in class lecture 5, 11 April. Continue reading, and solving the exercises, all the way to page 26. 5. Natural Transformations Definition 5.1. Let C and D be categories, and let F,G : C D be functors. A natural transformation, or a morphism of functors η : F G is a collection η = {η C } C Ob(C) of morphisms η C : F(C) G(C) in the category D. The condition is this: (N) For each morphism ϕ : C 1 C 2 in C there is equality η C2 F(ϕ) = G(ϕ) η C1 of morphisms in D. In other words, the diagram F(C 1 ) F (ϕ) F(C 2 ) η C1 G(C 1 ) G(ϕ) η C2 G(C 2 ) 23

Course Notes Amnon Yekutieli 25 April 2018 in D is commutative. Here is the diagram of functors: (5.2) C It is customary to draw morphisms of functors in such diagrams as doubled arrows. η F G Definition 5.3. In the situation of Definition 5.1, the morphism of functors η is called an isomorphism of functors if for every object C C the morphism in the category D is an isomorphism. D η C : F(C) G(C) Here are a few examples of morphisms of functors. Example 5.4. Continuing with Example 4.10, let A f B д C be K-ring homomorphisms. Then there is an isomorphism of functors η : Rest f Rest д Restд f of K-linear functors ModC Mod A. The isomorphism η N, for N ModC, is the identity on the underlying K-module. Exercise 5.5. Continuing with Exercise 4.11, let A be a central K-ring, and let ψ : M 1 M 2 be some fixed homomorphism in Mod A. We have K-linear functors defined by For each N Mod A there is a morphism F 1, F 2 : Mod A Mod K F i := Hom A (M i, ). η N : F 2 (N ) F 1 (N ), η N := Hom A (ψ, id N ) in Mod K. Prove that the collection of morphisms η = {η N } N Ob(ModA) is a morphism of functors η : F 2 F 1. Example 5.6. We continue with Exercise 4.8. So A is a nonzero ring, and we have the functors Free : Set Mod A, Free(X ) := F fin (X, A) and the forgetful functor The composed functor Forg : Mod A Set. Forg Free : Set Set sends a set X to the underlying set of the A-module F fin (X, A). 24

Course Notes Amnon Yekutieli 25 April 2018 For a set X let be the function We want to prove that η X : X F fin (X, A) η X (x) := δ x F fin (X, A). η = {η X } X Ob(Set) is a morphism of functors η : Id Set Forg Free. So we have to check that for every morphism д : X Y in Set, i.e. a function, the diagram (5.7) Id(X ) Id(д) Id(Y ) η X (Forg Free)(X ) (Forg Free)(д) η Y (Forg Free)(Y ) in Set is commutative. Let us translate this to a diagram with the actual objects and morphisms: X д Y η X F fin (X, A) η Y tr д F fin (X, A) Now for an element x X we have and (η Y д)(x) = η Y (д(x)) = δ д(x) (tr д η X )(x) = tr д (δ x ) = δ д(x). These are equal, so diagram (5.7) is indeed commutative. Exercise 5.8. Suppose we are given three categories C 1, C 2, C 3 ; pairs of functors and morphisms of functors F i,g i : C i C i+1 ; η i : F i G i. Give the formula for the composed morphism η 2 η 1 : (F 2 F 1 ) (G 2 G 1 ) of functors C 1 C 3, and prove that it is indeed a morphism of functors. This operation is called horizontal composition of morphisms of functors. Here are the relevant diagrams of functors: F 1 F 2 (5.9) C 1 η 1 C 2 η 2 C 3 G 1 G 2 25

Course Notes Amnon Yekutieli 25 April 2018 and (5.10) C 1 F 2 F 1 η 2 η 1 C 3 G 2 G 1 Exercise 5.11. Suppose we are given categories C 1 and C 2 ; three functors F,G, H : C 1 C 2 ; and morphisms of functors η : F G and θ : G H. For every object C C 1 define the morphism (θ η) C := θ C η C : F(C) H(C) in C 2. Show that the collection θ η = { (θ η) C }C Ob(C 1 ) is a morphism of functors θ η : F H. This operation is called vertical composition of morphisms of functors. Here are the relevant diagrams of functors: F (5.12) C 1 η θ G C 2 and H F (5.13) C 0 θ η H The operations from the the previous two exercises can be combinded. This is very confusing, but in the end quite elementary. The data is this: we are given three categories C 1, C 2, C 3 ; triples of functors and pairs of morphisms of functors F i,g i, H i : C i C i+1 ; C 1 η i : F i G i, θ i : G i H i. It can be shown (by a very tedious calculation) that the exchange property holds: (5.14) (θ 2 η 2 ) (θ 1 η 1 ) = (θ 2 θ 1 ) (η 2 η 1 ) 26

Course Notes Amnon Yekutieli 25 April 2018 as morphisms F 2 F 1 H 2 H 1 of functors C 1 C 3. This situation is shown in the next diagram. F 1 F 2 η 1 η 2 (5.15) C 1 G 1 C 2 G 2 C 3 θ 1 θ 2 H 1 H 2 Remark 5.16. All these opertions are explained in [Mac2, Section XII.3], as part of the discussion of the 2-category Cat of all U-categories. We do not need to know about 2- categories; all we need is to have a good understanding of the properties of morphisms of functors. The horizontal and vertical composition of morphisms of functors is made concrete in the next exercise. Exercise 5.17. Let A, B and C be rings. Let us introduce temporary notation for the category of A-B-bimodules: BiMod(A, B). (Later, after we talk about opposite rings and tensor products of rings, this category will be denoted by Mod A K B op.) We are given M i BiMod(A, B) and N i BiMod(B,C), for i = 1, 2, 3. In Exercise 4.17 we saw that this data gives rise to K-linear functors F i := Hom A (M i, ) : Mod A Mod B and G i := Hom B (N i, ) : Mod B ModC. We are also given homomorphisms ϕ i : M i M i+1 in BiMod(A, B) and ψ i : N i N i+1 in BiMod(B,C). (1) Show that ϕ i induces a morphism of functors η i : F i+1 F i, and ψ i induces a morphism of functors θ i : G i+1 G i. (2) Show that the homomorphism ϕ 2 ϕ 1 induces the morphism of functors η 1 η 2, and that the homomorphism ψ 2 ψ 1 induces the morphism of functors θ 1 θ 2. (3) Show that for every i the homomorphisms ϕ i and ψ i induce the morphism of functors θ i η i. Remark 5.18. After we learn about tensor products, we will see that there is an isomorphism of functors G i F i Hom A (P i, ), where P i := M i B N i BiMod(A,C). 27

And then there is a morphism ϕ i ψ i : P i P i+1, which induces the morphism of functors θ i η i. comment: Read to here for lecture 5. comment: Next topic: Equivalence of Categories 28

Lecture 6, 25 April 2018. Definition 6.1. A functor 6. Equivalence of Categories F : C D is called an isomorphism of categories if the functions and are all bijective. F Ob : Ob(C) Ob(D) F C1,C 2 : Hom C (C 1,C 2 ) Hom D ( F(C1 ), F(C 2 ) ) If F is an isomorphism of categories, then it has an inverse functor F 1 : D C, with the obvious formulas. The inverse F 1 is unique. It turns out that the notion of isomorphism of categories is too restrictive, and that we need a more relaxed notion: equivalence of categories. As we will see, the notion of equivalence of categories is analogous to the notion of homotopy equivalence of topological spaces. Example 6.2. Let X be the origin in the real plane R 2, and let Y be the closed unit disk in R 2. We give R 2 the standard metric topology, and X,Y have the subspace topologies. The inclusion f : X Y is not an isomorphism in Top; it is not even bijective. Yet f is a homotopy equivalence: the constant function д : Y X is a homotopy-inverse of f. There is equality д f = id X, and there is a homotopy f д id Y. In the categorical setting the role of homotopies is played by the morphisms of functors. Definition 6.3. A functor F : C D is called an equivalence of categories if there is a functor and isomorphisms of functors and G : D C, η : G F Id C ζ : F G Id D. The functor G is called a quasi-inverse of F. The situation is symmetric: the quasi-inverse is also an equivalence of categories. Let me give an important example. G : D C 29

Course Notes Amnon Yekutieli 25 April 2018 Example 6.4. Let D := Set fin be the category of finite sets. Define C to the full subcategory of D on the set of objects {S i } i N, where The functor S i := {1,..., i}. F : C D is the inclusion. We will prove that F is an equivalence. Since there is exactly one object from each isomorphism class in C, there is no choice in the definition of the quasi-inverse functor G : D C on objects. For a finite set S of cardinality i we must take G(S) := S i. What we need to choose an isomorphism in C. We make matters simple by choosing η S : G(S) S i (6.5) η Si := id Si : S i Si for S i C. We need to say what G does on morphisms. Given a morphism ϕ : S T in D we define G(ϕ) : G(S) G(T ) by This make the diagram diagram G(ϕ) := η 1 T ϕ η S. (6.6) G(S) G(ϕ) G(T ) η S S ϕ commutative. It is easy to see that G is a functor. Note that the action of G on morphisms is determined by our choice of isomorphisms η S. By our choice in (6.5) we see that G F = Id C, so we take the isomorphism of functors And the collection of morphisms T η T ζ := id : G F Id C. η := {η S } S D is an isomorphism of functors η : F G Id D. Proposition 6.7. If F : C D is an equivalence of categories, and if G,G : D C are both quasi-inverses of F, then there is an isomorphism of functors G G. Exercise 6.8. Prove this proposition. 30

Proposition 6.9. If F : C D is an equivalence of categories, then for every pair of objects C 1,C 2 C the function F : Hom C (C 1,C 2 ) Hom D ( F(C1 ), F(C 2 ) ) is bijective. Exercise 6.10. Try to prove this proposition. It is a bit tricky. If you can t, then there is a proof here: Solution 7.18. Another important example of an equivalence is in the next exercise. Exercise 6.11. Let K be a field, let D := Mod fin K be the category of finitely generated K-modules (aka finite dimensional vector spaces), and let C be the full subcategory of D on the set of objects {K i } i N. The functor F : C D is the inclusion. Prove that F is an equivalence. Remark 6.12. In the last example, let us view K i as a column module (for i > 0). Then the endomorphism ring of K i is the ring of matrices Mat i (K). If M Mod fin K is some rank i module, then its endomorphism ring is isomorphic to Mat i (K), and the isomorphism depends on our choice of quasi-inverse G : D C. Indeed, by Propositions 6.9 and 6.16, and Exercise 4.14, we have a K-ring isomorphism Definition 6.13. Let be a functor. G : End K (M) End K (K i ) = Mat i (K). F : C D (1) The functor F is called full (resp. faithful) if for every pair of objects C 1,C 2 C the function F : Hom C (C 1,C 2 ) Hom D ( F(C1 ), F(C 2 ) ) is surjective (resp. injective). (2) The functor F is called essentially surjective on objects if for every object D D there is an object C C with an isomorphism F(C) D in D. Theorem 6.14. Let F : C D be a functor. The following two conditions are equivalent. (i) F is an equivalence of categories. (ii) F is full, faithful and essentially surjective on objects. Exercise 6.15. Prove this theorem. (Hint: in Proposition 6.9 we saw that an equivalence F is full and faithful; so this is almost the implication (i) (ii). For the opposite implication try to imitate Example 6.4.) The next proposition was not in the lecture. 31

Course Notes Amnon Yekutieli 25 April 2018 Proposition 6.16. Let M and N be K-linear categories, and let F : M N be a K-linear functor. Assume that F is an equivalence, and G : N M is a quasi-inverse of F. Then G is a K-linear functor. Exercise 6.17. Prove this proposition. (Hint: see my proof of Proposition 6.9, i.e. Solution 7.18.) 7. Opposite Rings and Tensor Products Definition 7.1. Let A be a K-ring. The opposite ring of A is the K-ring A op, that has the same underlying K-module structure, but with multiplication The unit element remains the same. The identity function a 1 op a 2 := a 2 a 2. (7.2) op : A A op is a ring anti-isomorphism. We can view right A-modules as left op-modules. Indeed, given a right A-module M, define a left multiplication by elements of A op as follows: a op m := m a. It is clear that the unit element acts as the identity automorphism of M. As for associativity: (a 1 op a 2 ) op m = m (a 1 op a 2 ) = m (a 2 a 1 ) = (m a 2 ) a 1 = a 1 op (m a 2 ) = a 1 op (a 2 op m). We can make it very formal using our fancy language: Proposition 7.3. There is a K-linear isomorphism of categories such that the diagram F : (right A-modules) Mod A op (right A-modules) F Mod A op is commutative. Note that A = A op iff A is commutative. We now talk about tensor products. Forg Forg Mod K Definition 7.4. Let A be a K-ring, M Mod A op, N Mod A and P Mod K. An A-bilinear function β : M N P is a function with the following four properties: β(m 1 + m 2, n) = β(m 1, n) + β(m 2, n) β(m, n 1 + n 2 ) = β(m, n 1 ) + β(m, n 2 ) β(m a, n) = β(m, a n) 32

β(λ m, n) = β(m, λ n) = λ β(m, n) These must hold for every m,m i M; n, n i N ; a A and λ K. Example 7.5. If M = A and P = N, then β(a, n) := a n is an A-bilinear function. Definition 7.6. Let A be a K-ring, M Mod A op and N Mod A. A tensor product of M and N over A is a pair (P, β), where P Mod K, and β : M N P is an A-bilinear function. The pair (P, β) must have this universal property: (T) For every pair (P, β ) of this sort, there is a unique K-linear homomorphism ϕ : P P such that β = ϕ β. Theorem 7.7. Let A be a K-ring, M Mod A op and N Mod A. A tensor product (P, β) of M and N over A exists, and it is unique up to a unique isomorphism. Proof. Uniqueness: suppose (P, β) and (P, β ) are both tensor products of M and N over A. By property (T) there are unique homomorphisms ϕ : P P and ϕ : P P that interact with β and β as specified. The standard argument shows that ϕ and ϕ are inverse to each other. Existence: Let P be the free K-module on the set M N. Consider the K-submodule R P generated by these four types of elements: (m 1 + m 2, n) (m 1, n) (m 2, n) (m, n 1 + n 2 ) (m, n 1 ) (m, n 2 ) (m a, n) (m, a n) (λ m, n) λ (m, n) Define the K-module P := P/R and the function β : M N P, β(m, n) := (m, n) + R. The end of the proof is left as an exercise. Exercise 7.8. Finish the proof. (Hint: see proof of the commutative theorem, [Ye2, lecture 4, page 39].) Definition 7.9. The tensor product gets this notation: and The elements m n are called pure tensors. M A N := P m n := β(m, n). Proposition 7.10. The K-module M A N is generated by the pure tensors. Exercise 7.11. Prove this proposition. (Hint: study the proof of Theorem 7.7.) Proposition 7.12. Let ϕ : M 1 M 2 be a homomorphism in Mod A op, and letψ : N 1 N 2 be a homomorphism in Mod A. Then there is a unique homomorphism ϕ ψ : M 1 A N 1 M 2 A N 2 in Mod K, such that (ϕ ψ )(m n) = ϕ(m) ψ (n) 33